Stanford CME 295: A complete 16-hour curriculum on LLMs
techNmak · x · 2026-08-30
The author highly recommends Stanford's CME 295 course, suggesting it replaces half of the random LLM tutorials in your bookmarks. The 9-lecture, 16-hour course covers the modern LLM stack with excellent progression.
Syllabus highlights:
- Transformers: Tokenization to Attention
- Architectures: MHA/MQA/GQA, RoPE, BERT
- LLMs: MoE, Context Length, CoT
- Training: Pretraining, Quantization, SFT, LoRA
- Post-training: RLHF, PPO, DPO
- Reasoning: Inference-time scaling
- Agents: RAG, Function Calling, ReAct
- Evaluation: LLM-as-a-judge
- Current Directions: Multimodal & Diffusion
The course excels in ordering, helping build a clean mental model by connecting isolated concepts. Slides, cheatsheets, and exams with solutions are provided.
More from Companies & People
- When Will an AI Agent Autonomously Gain 1M Followers? — bilawalsidhu · 2026-08-30
- CIOs struggle with AI economics and agent governance — perilli · 2026-08-30
- ClaudeCode Creator Defines 5 New Product Roles for the AI Era — AGI Hunt · 2026-08-30
- All-In Podcast: China's Robot Olympics, Optimus, and AI Capex Narratives — CyberRobooo · 2026-08-30
- Replit exec on AI apps: Expertise and judgment are the moat when code is free — petergyang · 2026-08-30
- SEC filings this week: 50MW Microsoft compute deal, banks roll out agentic AI workspaces — Justgototheeffinmoon · 2026-08-30